一种基于特征点的无人机影像自动拼接方法  

A Automatic Mosaic Method in Unmanned Aerial Vehicle Images Based on Feature Points

作　　者：鲁恒[1] 李永树[1] 何敬[1] 陈强[1] 任志明[1] 

机构地区：[1]西南交通大学地理信息工程中心,四川成都610031

出　　处：《地理与地理信息科学》2010年第5期16-19,28,共5页Geography and Geo-Information Science

基　　金：国家"十一五"科技支撑计划项目(2006BAJ05A13)

摘　　要：论述了当前无人机影像快速拼接可选用的方法及可行性,将稳健的SIFT算法引入无人机影像自动拼接中,分析了该算法各阶段所消耗的时间。结合无人机自身的特点对算法进行改进,在进行特征点提取前通过估算相邻影像间的重叠度缩小了搜索范围;进行尺度空间极值点探测时通过实验获取了适应于无人机影像的最优高斯核尺寸,克服了传统SIFT算法采用固定核尺寸方法的缺陷,既减少了时间消耗,又尽可能多地获得了特征点,最后应用LM方法求得精确变换矩阵,完成了影像镶嵌。实验结果表明,该算法对无人机影像拼接具有较好的适应性,在保证算法鲁棒性的同时,提高了精度和效率。The methods which current unmanned aerial vehicle（UAV） phantom fast splicing may select and their feasibility were elaborated,the steady SIFT algorithm was introduced into the automatic mosaic of UAV images,and the time consumption of the algorithm during various stages was analyzed.Considering the feature of UAV to improve the algorithm,which can reduce the range for searching feature points through estimating the overlap of neighboring images,and the most superior size of Gaussian kernel adapting in the UAV images was obtained through the experiment when detecting the criterion space extreme points,it can overcome the traditional SIFT algorithm′s flaw by using the fixed size method,not only reduce the time consumption,but also obtain the characteristic points as far as possible.Finally the precise transformation matrix was obtained to complete images mosaic by using the LM method.Experimental results show that this algorithm has good compatibility to automatic mosaic of UAV images,which can remain the algorithm′s robust,meanwhile,the precision and the efficiency were increased.

关 键 词：无人机影像 自动拼接 影像重叠度 高斯核尺寸 

分 类 号：P231.5[天文地球—摄影测量与遥感]